Throttling a BCS crawl is done the same way as other SharePoint crawls, using crawler impact rules.
The main trick is figure out what site name to use. The most precise way is to use a combination of the LobSystem name and the LobSystemInstance name. In my case it was "recipeslob_recipesinst" :
An alternative is to use the LobSystem name with a wildcard (e.g. recipeslob_*).
Now we can set our rule (click on “Crawler Impact Rules” in the Crawling group of the Search Service Application management in Central Admin:
From there click “Add Rule” and fill it in:
To see that it is working, watch the Search Gatherer - SharePointServerSearch / Threads Accessing Network prefmon counter. In this screen shot you can see during the crawl only one thread accessing the network. Also observer that the requests in my console output are going in a nice neat linear order:
Now I’ll change the rule to 4 and run again:
Notice that now there are four threads accessing the network and that the requests are coming in bunches.
Yes, great tip !